We've improved our core tool interfaces and docs to simplify tool integrations and better handle diverse inputs, plus return complex outputs. These improvements enable more robust tool use in LangChain and reduce the manual effort of writing custom wrappers or interfaces.

They allow you to:

• Turn any code into a tool by passing any function into ChatModel.bind_tools()

• Determine which inputs should be generated by the model vs. other sources

• Enrich your tool outputs with an artifact attribute to use them in downstream actions

• Create fallbacks and handle tool calling errors to increase app reliability

Read more in the blog post: https://blog.langchain.dev/improving-core-tool-interfaces-and-docs-in-langchain/